Recherche Images Maps Play YouTube Actualités Gmail Drive Plus »
Connexion
Les utilisateurs de lecteurs d'écran peuvent cliquer sur ce lien pour activer le mode d'accessibilité. Celui-ci propose les mêmes fonctionnalités principales, mais il est optimisé pour votre lecteur d'écran.

Brevets

  1. Recherche avancée dans les brevets
Numéro de publicationUS5825907 A
Type de publicationOctroi
Numéro de demandeUS 08/893,668
Date de publication20 oct. 1998
Date de dépôt11 juil. 1997
Date de priorité28 déc. 1994
État de paiement des fraisPayé
Numéro de publication08893668, 893668, US 5825907 A, US 5825907A, US-A-5825907, US5825907 A, US5825907A
InventeursAnthony Peter Russo
Cessionnaire d'origineLucent Technologies Inc.
Exporter la citationBiBTeX, EndNote, RefMan
Liens externes: USPTO, Cession USPTO, Espacenet
Neural network system for classifying fingerprints
US 5825907 A
Résumé
A system for automatically classification of human fingerprints. An unidentified fingerprint is processed to produce a direction map. The direction map is processed to generate a course direction map. The coarse direction map is input to a locally connected, highly constrained feed-forward neural network. The neural network has a highly structured architecture well-suited to exploit the rotational symmetries and asymmetries of human fingerprints. The neural network classifies the unidentified fingerprint into one of five classifications: Whorl, Double Loop, Left Loop, Right Arch and Arch.
Images(9)
Previous page
Next page
Revendications(10)
What is claimed is:
1. A system for classifying an unknown fingerprint image having a plurality of pixels, comprising,
means for generating a pixel direction map from ridge directions found in the unknown fingerprint image;
means for generating a coarse direction map by dividing said pixel direction map into regions and determining for each of the regions a coarse direction as a most frequent pixel direction within the region; and
a neural network classifier having an input layer of neurons connected to receive said coarse direction map and connected to a hidden layer of neurons, wherein square regions of said input layer of neurons are arranged in a plurality or rings and connected to said hidden layer of neurons, said neural network classifier further includes an output layer of neurons that represent a plurality of fingerprint classifications.
2. The system of claim 1, wherein said neural network classifier classifies the unknown fingerprint image into either a left loop, a right loop, an arch, a whorl or a double whorl.
3. The system of claim 1, wherein said neural network classifier is a locally connected, highly constrained feed-forward neural network.
4. The system of claim 1, wherein said square regions within each of said plurality of rings overlap.
5. The system of claim 1, wherein global sections of said input layer of neurons are connected to said hidden layers of neurons, wherein said global sections are generated by dividing said coarse direction map into four equal quadrants.
6. The system of claim 1, wherein said input layer of neurons is connected to said hidden layer of neurons via a plurality of weights, each of said square regions within each of said plurality of rings from said input layer of neurons is connected via the same set of said plurality of weights to said hidden layer of neurons.
7. The system of claim 1, wherein said neural network classifier has at least two hidden layers.
8. The system of claim 7, wherein said at least two hidden layers include a first hidden layer connected to a second hidden layer using weight sharing.
9. The system of claim 8, wherein said neural network classifier evaluates the unknown fingerprint image from four different distances from the center of the unknown fingerprint image.
10. A system for classifying an unknown image having radial characteristics and a plurality of pixels, comprising,
means for generating a pixel direction map from ridge directions found in the unknown image;
means for generating a coarse direction map by dividing said pixel direction map into regions and determining for each of the regions a coarse direction as a most frequent pixel direction within the region; and
a neural network classifier having an input layer of neurons connected to receive said coarse direction map and connected to a hidden layer of neurons, wherein square regions of said input layer of neurons are arranged in a plurality or rings and connected to said hidden layer of neurons, said neural network classifier further includes an output layer of neurons that represent a plurality of fingerprint classifications.
Description

This is a Continuation of application Ser. No. 08/364,914 filed Dec. 28, 1994, now abandoned.

BACKGROUND OF THE INVENTION

1. Field of the Invention

This invention relates generally to the field of fingerprint identification and, more particularly, to a system and method for automatically classifying a human fingerprint using a locally connected, highly constrained feed-forward neural network, whereby the classified human fingerprint can then be readily identified.

2. Related Art

For over 100 years, fingerprints have been the most reliable means of identifying an individual; this is borne out by the fact that no two identifying fingerprints have ever been found. Conventionally, fingerprints have been identified by manually trying to match an unknown fingerprint against a set of known fingerprints. Recently, however, automatic identification of fingerprints has attracted considerable research attention.

Currently, commercial products exist that compare unidentified fingerprints, or sets of fingerprints, with established record files stored in a database. However, the comparison process is still not fully automatic. Additionally, limitations set by processing speed and false matchings prevent current systems from comparing unidentified fingerprints with each record in the database. Some degree of prior classification, and a corresponding organization of the record file, are needed in order to limit the number of automatic comparisons. The classification of the unidentified fingerprint is conventionally done manually.

The currently available commercial systems work adequately with work loads and database sizes typical at state and local agencies. However, at the national level the identification problem is compounded in two ways; the number of identifications requested each day, and the size of the database to be searched, are both much larger. A classification system groups fingerprints according to some visually distinguishable set of features. Classification of a fingerprint helps reduce search time in identifying unknown fingerprints. However, the need for a highly accurate automatic classifier still exists. For a general description of fingerprint identification see Russo, A. P. Automatic System For Fingerprint Analysis, Master's Thesis, Rensselaer Polytechnic Institute, August 1986.

Neural networks (NNs) have emerged as powerful tools for automatic pattern recognition tasks, especially those for which conventional methods fall far short of human performance. A neural network is a computing system made up of many simple, interconnected processing elements. It processes information by its dynamic state response to external inputs. Each processing element receives a number of inputs which are weighted according to their significance. From the weighted total input, the processing element computes a single output signal. In computing the output signal, each processing element learns, i.e., the weights on the inputs change in response to a given input and output pair.

Neural networks are suitable for automatic pattern recognition because the neural networks' ability to "learn" makes them ideal for handling nonlinear problems, which often involve data that are noisy or imprecise. But while theory suggests the broad utility of neural networks, the design of the network remains a difficult task. Many variables must be considered, such as defining the network topology and discovering the optimal set of weights for the network topology. The chosen network topology must hold sufficient complexity to address the given classification task.

Locally connected, highly constrained feed-forward neural networks of a different typology have been applied to Optical Character Recognition (OCR). See Y. LeCun, "Generalization and Network Design Strategies," in R. Pfeifer et al., editors, Connectionism in Perspective, pp. 143-156, Elesevier Science Publishers, Amsterdam, The Netherlands, 1989 and C. Crawford and E. Mjolsness, "Automated Fingerprint Identification: An Independent Study," Yale University, October 1991, where they are designated as convolutional and bipyramidal networks, respectively. However, these specific neural network architectures are not primarily designed for recognition of images having radial features (e.g., fingerprints).

SUMMARY OF THE INVENTION

The present invention provides a method for automatically classifying a human fingerprint using a neural network. The neural network is input a coarse direction map of the fingerprint, which describes the direction of the ridge flow in local regions of the print. The neural network has a highly structured architecture well-suited to exploit the rotational symmetries and asymmetries of human fingerprints. In accordance with a preferred embodiment, the neural network uses square regions of the coarse direction map arranged in rings to emphasize the radial characteristics of fingerprint patterns. The neural network is trained by back-propagation to classify the print into one of the following classes: Left Loop, Right Loop, Arch, Whorl (except for Double Loop), or Double Loop. These classes can then be used to reduce the number of fingerprints to be searched for identification requests.

BRIEF DESCRIPTION OF THE DRAWINGS

The foregoing and other features and advantages of the invention will be apparent from the following, more particular description of a preferred embodiment of the invention, as illustrated in the accompanying drawings.

FIG. 1 is a functional block diagram of the present invention.

FIG. 2 illustrates the hardware platform 200 used in a preferred embodiment of the present invention.

FIG. 3A through 3E show the architecture of the neural network's first layer of connections.

FIG. 4A shows an original fingerprint image before processing.

FIG. 4B shows a coarse direction map for the fingerprint image shown in FIG. 4A.

FIG. 5 shows the entire architecture of the neural network.

FIG. 6A through 6E show examples of an fingerprint in each of the classifications Arch, Whorl, Double Loop, Left Loop and Right Loop, respectively.

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

The preferred embodiment of the invention is discussed in detail below. While specific part numbers and/or configurations are discussed, it should be understood that this is done for illustration purposes only. A person skilled in the relevant art will recognize that other components and configurations may be used without parting from the spirit and scope of the invention.

The preferred embodiment of the invention is now described with reference to the figures where like reference numbers indicate like elements. Also in the figures, the left most digits of each reference number corresponds to the figure in which the reference number is first used.

FIG. 2 illustrates the hardware platform 200 used in a preferred embodiment of the present invention. It was chosen to ensure compatibility with, and a smooth transition to, a fully networked Automatic Fingerprint Identification System (AFIS). The hardware platform includes a Sun Sparc Station 10 220 with SuperSPARC 33 MHz uniprocessor (77.5 MIPS/46.6 SPECfp92), 128 Mbytes RAM, GX Adapter, 16" color monitor 270, mouse 290, keyboard 280, internal 424 Mbyte SCSI disk drive 222 and floppy disk drive 224. Monitor 270, keyboard 280, and mouse 290 are also provided to facilitate convenient data entry and display.

Speed and memory are required to quickly train a neural network. A stand alone disk subsystem 240 with two 2.9 Gbyte disk drives 242, 244 and 5.0 Gbyte EXABYTE 8 mm tape drive 246--the disk subsystem 240 will store fingerprint images to be processed. The high capacity tape drive 246 will allow convenient backup.

FIG. 1 is a functional block diagram of the present invention. Generally, a raw fingerprint image 110 is processed by a first processor module 115 to produce a pixel direction map 120. An exemplary raw fingerprint image is shown in FIG. 4A. The pixel direction map 120 is further processed by a second processor module 122 to produce a coarse direction map 130. An exemplary coarse direction map is shown in FIG. 4B. The course direction map 130 is used as an input to neural network classifier 135. The neural network classifier 135 outputs a fingerprint classification. In a preferred embodiment, the neural network classifier 135 classifies the original fingerprint 110 into one of the following classes: Left Loop, Right Loop, Arch, Whorl (except for Double Loop), or Double Loop. A fingerprint from each of these classifications is shown in FIGS. 6A through 6E, respectively.

The neural network architecture described herein can be used to identify other classes of fingerprints as well by modifying the inputs to neural network classifier 135 and/or neural network classifier 135 itself. Subclasses, like tented arches and central pocket loops, can be distinguished from plain arches and plain whorls, respectively, if the input coarse direction map 130 has enough resolution. Resolution can be increased, for example, at the expense of having more than 30×30 inputs to neural network classifier 135 (twice the resolution would imply 60×60 inputs). Resolution can also be increased by keeping only those 30×30 directions from the pixel direction map 120 that surround the fingerprint core (i.e., center). The fingerprint core can be identified by human or machine. Note that other areas of the fingerprint image 110 can be used as required and not just the fingerprint core.

The pixel direction map 120 (also referred to as a directional image) is created using the following equation: ##EQU1## where direction (i,j) is the direction at pixel (i,j), raw-- image(i,j) is the gray level value of the raw image at pixel (i,j), raw-- image(im,jm) is the gray level value of the raw image along the extrapolation in direction m, and L is the number of pixels used in computation. Typically, L=10 and m=16. A detailed discussion of the creation of pixel direction map 120 can be found in B. M Mehtre and B. Chatterjee, "Segmentation of Fingerprint Images--A Composite Method," Pattern Recognition, Vol. 22, No. 4, pp. 381-385, 1989. See also L. O'Gorman and J. Nickerson, "An Approach to Fingerprint Filter Design," Pattern Recognition, pp. 29-38, January 1989.

The algorithms for creating pixel direction map 120 assign to each pixel in the original fingerprint image 110 a direction that approximates the direction of ridges in the neighborhood of that pixel. Ridges are defined as the raised lines on the finger surface that form the pattern of the print.

The pixel direction map 120 is used to generate a coarse direction map 130. Essentially, the raw fingerprint image 110 of size M×N pixels (typically 512×512 or 768×832) is preprocessed to extract a coarse P×Q direction map 120 (P=30 and Q=30 in a preferred embodiment). Directions are angles and range from 0 to Pi (i.e., 3.14. . . ). To obtain a coarse direction map 130 of size P×Q, the pixel direction map 120 is divided into P×Q regions, implying that each region is of size R=M/P×S=N/Q, where R and S are rounded to the nearest integer value. Then for a given region of size R×S, the coarse direction in the coarse direction map 130 is simply the most frequent direction of all the pixel directions in the region. For example, given a pixel map region of size 10×10 pixels, where 51 of the pixels have direction D and the rest have some other direction, then the direction of the region in the coarse direction map would be D.

The step of transforming the pixel direction map 120 to a coarse direction map 130 provides an effective means of data reduction without loss of information important to classifying the fingerprint. Nevertheless, fingerprint images are noisy, and even the highest quality image processing will result in a coarse direction map with imperfections. To ensure accurate classification, the neural network classifier must be robust to these artifacts as well as translation, rotation and elastic deformation of the print. Accordingly, a convolutional neural network is used in a preferred embodiment (i.e., a locally connected, highly constrained feed-forward neural network). However, note that a fully-connected feedforward neural network or locally-connected feedforward neural network can also be used to implement the present invention.

Weight constraints in convolutional neural networks force the emergence of maps for feature extraction during the training cycle. Good neural network architecture design will incorporate specific knowledge of the problem to stimulate feature maps, and to incorporate smoothing layers to improve robustness to noise and image artifacts.

Besides accurate classification, there are other advantages to a convolutional neural network (see, e.g., A. P. Russo, "Tutorial Feedforward Neural Network Design for Signals and Images," Conference on Neural Networks for Ocean Engineering, Washington, D.C. (August 1991)), which include: constrained neural networks train faster and require fewer training samples; convolutional neural networks tend to be smaller and have fewer free parameters and connections; and convolutional neural network requires fewer training epochs, between 30 and 100, to reach peak generalization because of "hints" and prior knowledge built into the architecture.

A detailed description of the architecture used in a preferred embodiment of the present invention is set out below. Neural network classifier 135 has 30×30=900 input neurons which are fed the 30×30 directions from the coarse direction map 130. The actual value put into an input neuron is cos(direction), since cos(×) will range from -1 to +1 for the range of directions. This step ensures that the input neurons receive a smoothly defined, symmetric floating point data.

FIG. 5 illustrates a detailed architecture (including connectivity) of neural network 135. Neural network 135 has nine-hundred input neurons 510, sixty-eight neurons in a first hidden layer 520, twenty-four neurons in a second hidden layer 530, and five output neurons 540. The five output neurons 540 correspond to the five fingerprint classifications: Whorl, Double Loop, Left Loop, Right Loop, and Arch. In accordance with a preferred embodiment, neural network classifier 135 uses square regions of local connectivity arranged in rings to emphasize the radial characteristics of fingerprint patterns. This design provides a degree of rotational invariance.

FIG. 3A illustrates a first input architecture. Input neurons 510 are grouped into 16 square regions of size 6×6=36 pixels that are arranged to overlap and form Ring 1. Each group of thirty-six input neurons in Ring 1 is connected to a designated neuron in first hidden layer 520 via thirty-six different, independent weights. Each region in Ring 1, however, uses the same set of thirty-six weights for the connections to its respective hidden layer neuron in first hidden layer 520. This is the technique of weight sharing: during training of neural network 135, the set of thirty-six weights of a given square region in Ring 1 is forced to be the same for any other square region in that ring. That is, inputs to neural network 135 are constrained to perform the same operation on different parts of the image. Weight Sharing is discussed in a variety of references including Touretzky, Advances in Neural Information Processing Systems2, Carnegie Mellon University, Morgan Kaufmann Publishers, pp. 409-411 and D. E. Rumelhart, G. E. Hinton, R. J. Williams, "Learning Internal Representations by Error Propagation," Parallel Distributed Processing: Explorations in the Microstructure of Cognition, Vol. 1, MIT Press, Cambridge, Mass., 1986.

FIG. 3B illustrates a second input architecture. Ring 2 is formed in an analogous manner using sixteen square regions of size 4×4. Weight sharing is used in this ring as well. That is, the same set of sixteen weights are used for each region of Ring 2.

FIG. 3C illustrates a third input architecture. Ring 3 is formed from sixteen square regions of size 3×3. Again, weight-sharing is used for each group of nine weights.

FIG. 3D illustrates a fourth input architecture. Ring 4 is formed from 16 square regions of size 2×2. Again, weight-sharing is used for each group of four weights.

It is important to note that overlapping regions are used to form the rings; overlapping is used to make the neural network more robust against translation and rotation of the fingerprint image 110.

FIG. 3E illustrates a fifth input architecture. The nine hundred input neurons 510 are partitioned into four 15×15 squares or global sections. Again weight-sharing is used for each group of 225 weights. The global sections detect symmetries among the four quadrants of the image. They are used to impart knowledge of gross pattern differences.

Referring again to FIG. 5, regions of each ring are connected to a neuron in first hidden layer 520. In total, neural network classifier 135 has sixteen regions, each in Rings 1, 2, 3 and 4, plus four large rectangular sections. Thus, there must be 16+16+16+16+4=68 neurons in first hidden layer 520.

Neurons in second hidden layer 530 are fully connected to neurons in first hidden layer 520. That is, each neuron in second hidden layer 530 is connected to every neuron in first hidden layer 510. There are twenty-four neurons in second hidden layer 530; there is no weight-sharing used for these connections.

The output layer 540 is fully connected to second hidden layer 530. There are five output neurons, one for each pattern class. There is no weight-sharing used for these connections. The fingerprint is classified into one of the following classes: Left Loop, Right Loop, Arch, Whorl (except for Double Loop), or Double Loop.

The architecture of neural network 135 is designed to exploit the circular symmetry and asymmetry of the various fingerprint pattern types. The most important part of the neural network architecture is the connectivity from the input layer to the first hidden layer. Each ring evaluates the degree of symmetry of the fingerprint at its particular radius from the center. In particular, the collection of rings measures this symmetry at four different distances from the fingerprint image's center.

Generally, for a whorl pattern, the circular symmetry should hold for all four rings. For a double loop, this symmetry will hold for only the outer rings (i.e., Ring 1 and Ring 2). At all radii, left loops and right loops are fairly symmetric except for one quadrant (the lower left and lower right quadrants, respectively), so all four rings should easily detect the same asymmetry in the lower quadrant. Arches exhibit no circular symmetry but are rather symmetric across all four image quadrants; hence, the four large rectangular regions are used to detect these pattern types, where each rectangular region would detect the same (roughly) horizontal features. Note that, for obvious reasons, the four large rectangular regions are also helpful to detect the other pattern types.

Note that although two hidden layers are used in a preferred embodiment, other architectures can be implemented (e.g., one hidden layer, three hidden layer, etc) as should be readily apparent to those skilled in the art. Additionally, the present invention can be applied to images other than fingerprints. Essentially, the present invention can be implemented to recognize or classify images that have radial features.

The present invention is well suited to classify other types of images (not just fingerprints) with radial features. An example would be machine vision applications, where a robot might need to identify parts (e.g., screws and washers) from a bin. Gray-level images of those parts, providing the images are not too large in size, can be fed directly into neural network classifier 135 instead of the coarse map 130. In that case, the actual gray levels are input into neural network 135 instead of the cosine of the direction. While the neural network architecture might need to be modified to optimize performance for different images, the ring-like topology described above would produce a higher recognition rate.

Weight-sharing is used to minimize the number of free parameters that describe neural network 135. It is well known that neural networks with too many free parameters do not perform as well as, need more training exemplars, and take longer to train than neural networks with fewer free weights.

Training is accomplished using a standard backpropagation algorithm for feed-forward neural networks. Backpropagation is described in detail in D. E. Rumelhart, G. E. Hinton, R. J. Williams, "Learning Internal Representations by Error Propagation," Parallel Distributed Processing: Explorations in the Microstructure of Cognition, Vol. 1, MIT Press, Cambridge, Mass., 1986. In a preferred embodiment, a learning rate of 0.3 and a momentum term of 0.2 is used. This, of course, required a database of fingerprints whose true pattern class was labeled beforehand (by FBI experts), as backpropagation requires knowing what the activation level of each output neuron should be for each training example. (Activation level refers to the value of the output of a neuron.) In a preferred embodiment, during training, the output neuron corresponding to the true pattern class of the current example has a desired activation level of +1 and all other neurons have a desired level of -1. Therefore, it follows that a test example would be classified as the pattern corresponding to the output neuron with the highest activation level.

It should be understood that embodiments of the present invention can be implemented in hardware, software or a combination thereof. In such embodiment, the various component and steps would be implemented in hardware and/or software to perform the functions of the present invention. Any presently available or future developed computer software language and/or hardware components can be employed in such embodiments of the present invention.

While the invention has been particularly shown and described with reference to preferred embodiments thereof, it will be understood by those skilled in the art that various changes in form and details may be made therein without departing from the spirit and scope of the invention.

Citations de brevets
Brevet cité Date de dépôt Date de publication Déposant Titre
US5140642 *23 avr. 199118 août 1992Wen Hsing HsuMethod and device for allocating core points of finger prints
US5276771 *27 déc. 19914 janv. 1994R & D AssociatesRapidly converging projective neural network
US5465303 *12 nov. 19937 nov. 1995Aeroflex Systems CorporationAutomated fingerprint classification/identification system and method
US5519785 *31 août 199421 mai 1996Nec CorporationCorrecting method for directional data of streaked patterns and information processing apparatus for executing it and correcting method for pitch data of streaked patterns and information processing apparatus for executing it
US5572597 *6 juin 19955 nov. 1996Loral CorporationFingerprint classification system
US5659626 *20 oct. 199419 août 1997Calspan CorporationFingerprint identification system
Citations hors brevets
Référence
1 *A.P. Russo, An Automatic System for Fingerprint Analysis , Master s Thesis, Rensselaer Polytechnic Institute, May 1986.
2A.P. Russo, An Automatic System for Fingerprint Analysis, Master's Thesis, Rensselaer Polytechnic Institute, May 1986.
3 *Crawford et al., Automated Fingerprint Identification: An Independent Study , Yale University, Oct. 1991.
4Crawford et al., Automated Fingerprint Identification: An Independent Study, Yale University, Oct. 1991.
5 *D.S. Touretzky, Advances in Neural Information Processing Systems 2 , Carnegie Mellon University, Morgan Kaufmann Publishers, pp. 405 411.
6D.S. Touretzky, Advances in Neural Information Processing Systems 2, Carnegie Mellon University, Morgan Kaufmann Publishers, pp. 405-411.
7Mehtre et al. "Segmentation of Fingerprint Images--A composite Method," Pattern Recognition, vol. 22, No. 4, pp. 381-385, 1989.
8 *Mehtre et al. Segmentation of Fingerprint Images A composite Method, Pattern Recognition , vol. 22, No. 4, pp. 381 385, 1989.
9Rumelhart et al., "Learning Internal Representations by Error Propagation," Parallel Distributed Processing: Explorations in the Microstructure of Cognition, vol. 1, MIT Press, Cambridge, MA, 1985.
10 *Rumelhart et al., Learning Internal Representations by Error Propagation, Parallel Distributed Processing: Explorations in the Microstructure of Cognition , vol. 1, MIT Press, Cambridge, MA, 1985.
Référencé par
Brevet citant Date de dépôt Date de publication Déposant Titre
US623674119 févr. 199722 mai 2001Stmicroelectronics S.R.L.Method and device for identifying fingerprints
US633034728 juil. 199811 déc. 2001Stmicroelectronics S.R.L.Method and device for identifying fingerprints using an analog flash memory
US647224628 déc. 199929 oct. 2002Stmicroelectronics, Inc.Electrostatic discharge protection for integrated circuit sensor passivation
US6480617 *4 avr. 200112 nov. 2002Stmicroelectronics S.R.L.Method and device for identifying fingerprints
US649044331 août 20003 déc. 2002Automated Business CompaniesCommunication and proximity authorization systems
US6496021 *18 janv. 200217 déc. 2002Stmicroelectronics, Inc.Method for making a capacitive distance sensor
US651238129 déc. 200028 janv. 2003Stmicroelectronics, Inc.Enhanced fingerprint detection
US661055528 nov. 200026 août 2003Stmicroelectronics, Inc.Selectively doped electrostatic discharge layer for an integrated circuit sensor
US6920435 *10 avr. 200219 juil. 2005Indivos CorporationTokenless biometric electronic transactions using an audio signature to identify the transaction processor
US696145326 nov. 20011 nov. 2005Secugen CorporationMethod for extracting fingerprint feature data using ridge orientation model
US719716812 juil. 200227 mars 2007Atrua Technologies, Inc.Method and system for biometric image assembly from multiple partial biometric frame scans
US723922730 déc. 19993 juil. 2007Upek, Inc.Command interface using fingerprint sensor input system
US725658925 avr. 200214 août 2007Atrua Technologies, Inc.Capacitive sensor system with improved capacitance measuring sensitivity
US725957322 mai 200221 août 2007Atrua Technologies, Inc.Surface capacitance sensor system using buried stimulus electrode
US731043226 juin 200418 déc. 2007Artinnet Corp.Ported system for personal identity verification
US7471812 *21 mars 200530 déc. 2008Lg Electronics Inc.Fingerprint symmetry axis extraction method and fingerprint recognition method using fingerprint symmetry axis
US747477221 juin 20046 janv. 2009Atrua Technologies, Inc.System and method for a miniature user input device
US750561310 juil. 200617 mars 2009Atrua Technologies, Inc.System for and method of securing fingerprint biometric systems against fake-finger spoofing
US751955817 sept. 200214 avr. 2009Ballard Claudio RBiometrically enabled private secure information repository
US75870724 août 20048 sept. 2009Authentec, Inc.System for and method of generating rotational inputs
US769772930 juin 200413 avr. 2010Authentec, Inc.System for and method of finger initiated actions
US769856728 déc. 200513 avr. 2010Yt Acquisition CorporationSystem and method for tokenless biometric electronic scrip
US775159516 févr. 20076 juil. 2010Authentec, Inc.Method and system for biometric image assembly from multiple partial biometric frame scans
US780960117 oct. 20015 oct. 2010Johnson & Johnson Consumer CompaniesIntelligent performance-based product recommendation system
US783107018 févr. 20059 nov. 2010Authentec, Inc.Dynamic finger detection mechanism for a fingerprint sensor
US788203217 août 20001 févr. 2011Open Invention Network, LlcSystem and method for tokenless biometric authorization of electronic communications
US78854365 juil. 20078 févr. 2011Authentec, Inc.System for and method of assigning confidence values to fingerprint minutiae points
US794024931 oct. 200610 mai 2011Authentec, Inc.Devices using a metal layer with an array of vias to reduce degradation
US797067814 avr. 200928 juin 2011Lapsley Philip DBiometric financial transaction system and method
US8032473 *31 juil. 20084 oct. 2011Rice Daniel MGeneralized reduced error logistic regression method
US811549711 nov. 200814 févr. 2012Authentec, Inc.Pixel sensing circuit with common mode cancellation
US82310563 avr. 200631 juil. 2012Authentec, Inc.System for and method of protecting an integrated circuit from over currents
US826071628 déc. 20054 sept. 2012Open Invention Network, LlcSystem and method for processing tokenless biometric electronic transmissions using an electronic rule module clearinghouse
US8396258 *25 sept. 200712 mars 2013Hitachi Ltd.Finger vein pattern inputting device
US842189015 janv. 201016 avr. 2013Picofield Technologies, Inc.Electronic imager using an impedance sensor grid array and method of making
US845268028 oct. 201128 mai 2013Open Invention Network, LlcBiometric financial transaction system and method
US863093216 août 201214 janv. 2014Open Invention Network, LlcBiometric financial transaction system and method
US863093310 avr. 201314 janv. 2014Open Invention Network, LlcBiometric financial transaction system and method
US866684422 juin 20104 mars 2014Johnson & Johnson Consumer CompaniesIntelligent performance-based product recommendation system
US879179221 juin 201029 juil. 2014Idex AsaElectronic imager using an impedance sensor grid array mounted on or about a switch and method of making
US883199422 déc. 20109 sept. 2014Open Invention Network, LlcSystem and method for tokenless biometric authorization of electronic communications
US886634727 mai 201121 oct. 2014Idex AsaBiometric image sensing
US895884623 août 200617 févr. 2015Charles Freeny, IIICommunication and proximity authorization systems
US916532310 déc. 201220 oct. 2015Open Innovation Network, LLCBiometric transaction system and method
US923014914 sept. 20125 janv. 2016Idex AsaBiometric image sensing
US926898814 sept. 201223 févr. 2016Idex AsaBiometric image sensing
US960070413 mars 201321 mars 2017Idex AsaElectronic imager using an impedance sensor grid array and method of making
US965920819 nov. 201523 mai 2017Idex AsaBiometric image sensing
US978533013 févr. 200910 oct. 2017Apple Inc.Systems for and methods of providing inertial scrolling and navigation using a fingerprint sensor calculating swiping speed and length
US979891721 nov. 201624 oct. 2017Idex AsaBiometric sensing
US20020111917 *10 avr. 200215 août 2002Indivos Corporation, A Delaware CorporationTokenless biometric electronic transactions using an audio signature to identify the transaction processor
US20020161664 *17 oct. 200131 oct. 2002Shaya Steven A.Intelligent performance-based product recommendation system
US20030013328 *22 mai 200216 janv. 2003Andrade Thomas L.Connection assembly for integrated circuit sensors
US20030020495 *22 mai 200230 janv. 2003Andrade Thomas L.Surface capacitance sensor system using buried stimulus electrode
US20030101348 *12 juil. 200229 mai 2003Russo Anthony P.Method and system for determining confidence in a digital transaction
US20030115490 *12 juil. 200219 juin 2003Russo Anthony P.Secure network and networked devices using biometrics
US20030225693 *17 sept. 20024 déc. 2003Data Treasury CorporationBiometrically enabled private secure information repository
US20050041885 *4 août 200424 févr. 2005Russo Anthony P.System for and method of generating rotational inputs
US20050169503 *30 juin 20044 août 2005Howell Mark J.System for and method of finger initiated actions
US20050188213 *23 févr. 200425 août 2005Xiaoshu XuSystem for personal identity verification
US20050207625 *21 mars 200522 sept. 2005Lg Electronics Inc.Fingerprint symmetry axis extraction method and fingerprint recognition method using fingerprint symmetry axis
US20050286761 *26 juin 200429 déc. 2005Xiaoshu XuPorted system for personal identity verification
US20060147096 *4 janv. 20066 juil. 2006Dong-Jae LeeFingerprint region segmenting apparatus, directional filter unit and methods thereof
US20060229932 *7 mars 200612 oct. 2006Johnson & Johnson Services, Inc.Intelligent sales and marketing recommendation system
US20070014443 *10 juil. 200618 janv. 2007Anthony RussoSystem for and method of securing fingerprint biometric systems against fake-finger spoofing
US20070098228 *31 oct. 20063 mai 2007Atrua Technologies, IncDevices using a metal layer with an array of vias to reduce degradation
US20070207681 *3 avr. 20066 sept. 2007Atrua Technologies, Inc.System for and method of protecting an integrated circuit from over currents
US20070274575 *16 févr. 200729 nov. 2007Russo Anthony PMethod and system for biometric image assembly from multiple partial biometric frame scans
US20080013808 *5 juil. 200717 janv. 2008Russo Anthony PSystem for and method of assigning confidence values to fingerprint minutiae points
US20080075332 *25 sept. 200727 mars 2008Yasunori FujisawaFinger vein pattern inputting device
US20090089228 *31 juil. 20082 avr. 2009Rice Daniel MGeneralized reduced error logistic regression method
US20090123039 *11 nov. 200814 mai 2009Upek, Inc.Pixel Sensing Circuit with Common Mode Cancellation
US20090132445 *12 mai 200821 mai 2009Rice Daniel MGeneralized reduced error logistic regression method
US20100262556 *22 juin 201014 oct. 2010Johnson & Johnson Consumer Companies, Inc.Intelligent performance-based product recommendation
WO2016090439A1 *28 août 201516 juin 2016Universidade Estadual De Campinas - UnicampMethod for detecting the brightness of fingerprints using convolutional networks
Classifications
Classification aux États-Unis382/124, 382/224, 706/20, 382/156, 382/125, 382/170, 706/31
Classification internationaleG06K9/00
Classification coopérativeG06K9/0008
Classification européenneG06K9/00A2T
Événements juridiques
DateCodeÉvénementDescription
20 janv. 1998ASAssignment
Owner name: SILICON VALLEY BANK, CALIFORNIA
Free format text: SECURITY AGREEMENT;ASSIGNOR:VERIDICOM, INC.;REEL/FRAME:008953/0431
Effective date: 19971215
19 mars 1999ASAssignment
Owner name: TRANSAMERICA BUSINESS CREDIT CORPORATION, ILLINOIS
Free format text: SECURITY INTEREST;ASSIGNOR:VERIDICOM, INC.;REEL/FRAME:009822/0619
Effective date: 19990305
5 avr. 2001ASAssignment
Owner name: THE CHASE MANHATTAN BANK, AS COLLATERAL AGENT, TEX
Free format text: CONDITIONAL ASSIGNMENT OF AND SECURITY INTEREST IN PATENT RIGHTS;ASSIGNOR:LUCENT TECHNOLOGIES INC. (DE CORPORATION);REEL/FRAME:011722/0048
Effective date: 20010222
28 mars 2002FPAYFee payment
Year of fee payment: 4
22 mars 2006FPAYFee payment
Year of fee payment: 8
6 déc. 2006ASAssignment
Owner name: LUCENT TECHNOLOGIES INC., NEW JERSEY
Free format text: TERMINATION AND RELEASE OF SECURITY INTEREST IN PATENT RIGHTS;ASSIGNOR:JPMORGAN CHASE BANK, N.A. (FORMERLY KNOWN AS THE CHASE MANHATTAN BANK), AS ADMINISTRATIVE AGENT;REEL/FRAME:018590/0047
Effective date: 20061130
22 oct. 2008ASAssignment
Owner name: VERIDICOM INC, CALIFORNIA
Free format text: RELEASE;ASSIGNOR:SILICON VALLEY BANK;REEL/FRAME:021709/0939
Effective date: 20081020
30 oct. 2008ASAssignment
Owner name: VERIDICOM, INC, CALIFORNIA
Free format text: RELEASE BY SECURED PARTY;ASSIGNOR:GE TECHNOLOGY FINANCE CORPORATION SUCCESSOR-IN-INTEREST TO TRANSAMERICA BUSINESS CREDIT CORPORATION;REEL/FRAME:021785/0896
Effective date: 20081030
14 avr. 2010FPAYFee payment
Year of fee payment: 12